Intentionality of Act and the Future of Observations

Resolution

MnM call 20060605:

• Motion (Charlie McCay): This INM request for guidance (as of 20060605, see history page) will become M&M harmonization proposal for the next Harmonization meeting.
• Second(Mead Walker): 10-0-0 (passed)

Harmonization Action

• Update the definition of Act to clarify the above and remove "intentionality" from the definition of Act.

• Add guidance explaining the difference between SUBJ and VALUE

Background

There were discussions at the last few HL7 modeling harmonization events regarding a revision on the way we think about acts.

Here is a fragment of the Act Definition:

Acts are the pivot of the RIM; all domain information and processes are represented primarily in Acts. Any profession or business, including healthcare, is primarily constituted of intentional actions, performed and recorded by responsible actors. An Act-instance is a record of such an intentional action. Intentional actions are distinguished from something that happens by forces of nature (natural events). Such natural events are not Acts by themselves, but may be recorded as observed (Observation).

Problem

Some questions that came up are:

• Is an accidental administration of insulin to a patient a SubstanceAdministration, can it be called an "intentional action"?

• Is a snake bite or bee sting a SubstanceAdministation, can it be called an "intentional action"?

Snake bites and bee stings are "intentional acts." The bee decides to give its life to defend the hive, and the snake not only decides to bite, but to envenomate, as well. ...pete gilbert

• Is a gun-shot bullet a SubstanceAdministration?

• An abdominal surgecy is a Procedure, is someone stabbing a knife performing a procedure? Is it an intentional action?

• Is a car accident an intentional action?

• When a child swallows the contents of a bottle of poison is that an intentional action?

The Proposal

The Proposal has been to remove the requirement for Acts to be called intentional, and instead say that a snake bite can be called a substance administration, someone stabbing me with a knife is performing an Act, and a gun-shot hitting me in a cross-fire is also an Act. Car accident is an act as well and so is me being hit by a lightning strike.

So, all distinctions between acts that are ordered scheduled and billed for vs. those events of nature would go away. When the Act class was still called "Service" this kind of change was unthinkable, now it is a notion that is seriously entertained.

Rationale

The reason why people want this change is that it allows them to use the Act attributes and participations (time, performer, etc.) to describe any incident just like acts.

Points of Concern

The downside of this is that it causes a great shift in the modeling of Observations. In the past, people would use Observations to describe events of nature, and unintended incidents such as gunshot wounds or car accident injuries. In fact there is still a sizable group of people who would like to think of procedure history (e.g., appendectomy 20 years ago) as an Observation in a collection of "patient history" Observations. I think the RIM was trying to make that appendectomy a Procedure regardless of whether it is recent and known from first hand reports or if it is only known through patient recollection or even hearsay. But now we would be moving even more information items that used to be understood as Observations over into being other specific Acts such as Procedures or SubstanceAdministrations.

Considerations

If we are to move forward and allow snakebites, gunshots and swallowing poison to be fully represented as Acts, rather than just Observations, we need to come up with some guidelines. Specifically, we need to answer the following questions:

1. Are there some things that should always be modeled as Acts, and never Observations (keeping in mind that Observations are a type of Act)?

2. Are there some things that should be modeled as Observations and never as discrete Acts?

3. Are there some things that can appropriately be modeled either way, and if so, what's the criteria for deciding which approach to use?

4. Might there be circumstances where we would want both an Observational and an 'Act' view of the same event in an instance, and if so, how do we model the relationship between them?

Possible Guidance

• Observations related to the gathering and recording of information about past occurrences should generally be captured as discrete acts representing those occurrences, rather than observations

• The author and informer attributes of the Act can convey some of the information about who is making the "observation" by reporting the action

• Where the desire is to capture a primary action of "gathering and recording data", Observation may be used, even if the information gathered is historical. Doing so allows additional details about the actual "observation" such as supervisor, activityTime, methodCode, etc. that cannot be conveyed by simply stating the historical "Act" with an author and an informer.

• A new ActRelationship, constrained to having a source of type Observation, should be added that allows the conveying of "Value". Thus an Observation can convey it's value in the Value attribute (when you're just capturing a free text report of multiple Acts and don't have enough information to differentiate or expose them as separate acts) or by creating child Acts related through the "value" ActRelationship

Follow-up

2011-01-12: Lloyd will do the harmonization proposal to add ActRelationshipType#VALUE